deepmi/fastsurfer

Fastsurfer

FastSurfer is a fast and accurate deep-learning pipeline for the analysis of human brain MRI. FastSurfer provides a fully compatible FreeSurfer alternative for volumetric and surface-based thickness analysis, also supporting sub-mm resolutions, and sub-segmentation of neuroanatomical structures such as the cerebellum and hypothalamus.

Download

The easiest is to use Singularity:

singularity build fastsurfer-gpu.sif docker://deepmi/fastsurfer:latest

or Docker:

docker pull deepmi/fastsurfer:latest

You can find documentation, usage and install instructions here: [***]

Tags

cu###-v#.#.#: uses cuda (e.g. cu124 for 12.4) and FastSurfer with the specified versions. Older cuda versions are available to support systems with older Nvidia drivers. These also support CPU-only processing.

cpu-v#.#.#: CPU-only (smaller image without GPU support).

rocm#.#-v#.#.#: uses rocm for AMD GPU support with the specified versions (experimental).

Convenience Tags

These point to one of the images above, for convenience:

latest, gpu-latest: point to the latest FastSurfer version with latest cuda package.

cpu-latest: points to the latest CPU-only image.

cuda-v#.#.#: points to latest cuda version with specified FastSurfer version.

rocm-v#.#.#: points to latest rocm version with specified FastSurfer version.

Overview

Based on a T1-weighted MRI you get everything you need for quick structure localization, whole brain segmentation, extraction of quantitative measures, group analysis of your cohort, or structural pre-processing for fMRI/diffusion analysis:

• Deep-learning based whole brain segmentation into 95 classes in 1-4 minutes on the GPU, and 20 min on the CPU

  • Desikian-Killiany-Tourville atlas segmentation (=33 subcortical structures and 31 cortical structures per hemisphere)
  • Cerebellar lobe segmentation into 27 structures
  • Hypothalamus sub-segmentation
  • ROI-wise volume statistics
  • Supports native high-resolution segmentation (< 1.0mm^3, no resampling)

• Full FreeSurfer-conform outputs in approximately 45 min (+ 30 min for spherical registration, on by default)

  • Cortical surfaces (white, pial)
  • Surface measures (thickness, curvature and more)
  • Surface labels and annotations (aparc.annot, cortex.label, ...)
  • Point-wise surface statistics and ROI-wise volume statistics
  • Spherical registration to FSAVERAGE - directly usable for inter-subject registration and alternative atlas registration (e.g. Destrieux, Desikian-Killiany)
  • Supports native high-resolution MRIs (< 1.0mm^3, no resampling)

Check-out our https://colab.research.google.com/github/Deep-MI/FastSurfer/blob/stable/Tutorial/Tutorial_FastSurferCNN_QuickSeg.ipynb for a first taste of FastSurfer and generate your first FastSurfer segmentation in just three clicks!

Docker images

We provide images to run FastSurfer. You can choose between running the segmentation on the gpu or cpu. The Surface module depends on FreeSurfer. Thus, the packages contain a stripped-down FreeSurfer distribution (excluding available subjects and visualization tools). However, it does not include the valid license file needed to run FreeSurfer. You can either mount your license file every time you run this docker or point to it with the FS_LICENSE env variable (export before executing the docker run command). Obtain the FreeSurfer License https://surfer.nmr.mgh.harvard.edu/registration.html if you plan to run the surface module. For the segmentation modules no license is needed.

Links

Github Source Code: https://github.com/deep-mi/FastSurfer

FastSurfer Introduction: [***]

FastSurfer Documentation: [***]

FreeSurfer: https://surfer.nmr.mgh.harvard.edu

License

Apache License v2

References

If you use this for research publications, please cite:

Henschel L, Conjeti S, Estrada S, Diers K, Fischl B, Reuter M. FastSurfer - A fast and accurate deep learning based neuroimaging pipeline. NeuroImage 219 (2020), . https://doi.org/10.1016/j.neuroimage.2020.

Henschel L, Kügler D, Reuter M. FastSurferVINN: Building resolution-independence into deep learning segmentation methods - A solution for HighRes brain MRI. NeuroImage 251 (2022), . https://doi.org/10.1016/j.neuroimage.2022.

Faber J*, Kuegler D*, Bahrami E*, et al. (*co-first). CerebNet: A fast and reliable deep-learning pipeline for detailed cerebellum sub-segmentation. NeuroImage 264 (2022), . https://doi.org/10.1016/j.neuroimage.2022.

Estrada S, Kügler D, Bahrami E, et al. FastSurfer-HypVINN: Automated sub-segmentation of the hypothalamus and adjacent structures on high-resolutional brain MRI. Imaging Neuroscience (2023), 1:1-32. https://doi.org/10.1162/imag_a_00034